Table 3 Performance comparison results of different text automatic generation models.
From: Automatic generation and transfer of Yongju opera language using AIGC technology and style mapping
Model | Rise time | User satisfaction | BLEU | |||
|---|---|---|---|---|---|---|
Relevance | Fluency | Consistency | Overall satisfaction | |||
TFCPG | 33.32 ± 2.15** | 4.32 | 3.87 | 4.06 | 4.08 | 45.55 ± 1.32** |
Transformer | 76.81 ± 5.63 | 3.28 | 3.41 | 3.18 | 3.29 | 22.60 ± 2.41 |
PLM | 64.83 ± 4.94 | 2.54 | 3.62 | 3.33 | 3.16 | 25.38 ± 1.89 |
GPT | 55.26 ± 3.74 | 3.71 | 3.47 | 3.68 | 3.62 | 26.97 ± 2.18 |
BERT | 58.94 ± 4.25 | 3.63 | 3.79 | 3.41 | 3.61 | 30.16 ± 1.95 |
